Identification and analysis of putative virulence factors 1. Tyramine

Kanbar et al. (2004) showed that tyramine production of E. faecalis has highly toxic effects on honey bee larvae (Kanbar et al., 2005). Furthermore, the development of tyramine treated honey bee brood was impaired and showed classical EFB symptoms as treated larvae changed their color to yellow/brown. We identified an Enterococcus-type tyrosine decarboxylase gene cluster, which is involved in tyramine production (Connil et al., 2002).

Interestingly, the genes encoding the tyrosine decarboxylase of the typical strains 82, 90.0, 119 and ATCC 35311 are putatively dysfunctional due to a nonsense mutation (Supplementary Data 2, Sheet 4).

3.3.2. Bacteriocins

A high number of bacteria produce peptides called bacteriocins, which possess antimicrobial activities against very close related species or even against strains of the same species (Zacharof and Lovitt, 2012). A total of seven (typical strains) or five (atypical strain) genes and gene clusters encoding for putative bacteriocin biosynthesis and transport functions were identified in the genomes of the typical strains and the atypical strain DAT561 (Supplementary Data 2, Sheet 4). These clusters share high similarity to putative bacteriocin biosynthesis clusters of Enterococcus and transport clusters of Streptococcus spp.

(Supplementary Figure 1). Here, we only focus on putative functional genes encoding for bacteriocin biosynthesis proteins, although the functionality of most of the gene clusters is uncertain due to nonsense mutations in the corresponding ORFs (Supplementary Data 2, Sheet 4). Two putative functional ORFs share low similarity with Zoocin A-like bacteriocins (Supplementary Data 2, Sheet 4, see “Bacteriocin-associated proteins”, ORF1 and 2), and one with an unclassified bacteriocin determined by BAGEL3 (van Heel et al., 2013) (Supplementary Data 2, Sheet 4, see “Bacteriocin-associated proteins”, ORF3). ORF1 and ORF2 share 14% and 18% amino acid sequence identity (31% and 39% coverage) to Zoocin A, respectively, a streptococcolytic enzyme with weak β-lactamase activity (Heath et al., 2004). Remarkably, ORF1 is only present in typical strains of ST3 and ST7. On the contrary, ORF3 was found in all other ST determined in this study, excluding the atypical strain. Thus, putative bacteriocin biosynthesis proteins were only identified in typical strains. In addition, we found lysozyme subfamily 2 domain/GH73 family domain-containing proteins (Supplementary Data 2, Sheet 4), which might be involved in bacterial cell wall degradation (Joris et al., 1992).

3.3.3. Larval glycoprotein and peritrophic matrix degrading enzymes

The peritrophic matrix lines the midgut of invertebrates and is comprised of secreted chitin and (glyco)proteins, mainly peritrophins (Terra, 2001). It compartmentalizes digestive processes, protects from ingested xenobiotics, and acts as a mechanical barrier against abrasive food pieces and pathogens (Garcia-Gonzalez et al., 2014c; Terra, 2001). In the genomes of M. plutonius a potential chitin-binding domain-containing protein, consisting of

a signal peptide and a type 3 chitin binding domain, was identified (Supplementary Data 2, Sheet 4). It belongs to the auxiliary activity 10, a family of lytic polysaccharide monooxygenases, and exhibited 37% amino acid sequence similarity to PlCBP49 (JX185746) of P. larvae. PlCBP49 represents a key virulence factor of P. larvae and is able to degrade the peritrophic matrix of the honey bee larva (Garcia-Gonzalez et al., 2014c).

Additionally, a peptidase M60 family protein (enhancin), which can potentially degrade the peritrophic matrix (Toprak et al., 2012; Tellam et al., 1999; Fang et al., 2009; Peng et al., 1999) of the honey bee larvae, was detected (Figure 4, Figure 5, and Supplementary Data 2, Sheet 4). It contains a signal peptide and shows high similarity to an enhancin-like protein of Bacillus thuringiensis serovar kurstaki str. YBT1520 (51% identity, ACN22337) (Figure 5). The latter was shown to disrupt the insect midgut peritrophic matrix (Fang et al., 2009).

The peptidase M60 family protein also shows low amino acid sequence similarity to a M60 family protein of P. larvae DSM 25719 (22% identity, ERIC1_1c29890) (Djukic et al., 2014). Additionally, it is homologous to several pseudogenes of P. larvae DSM 25719 and P. larvae DSM 25430, which are fragmented by transposase insertions or mutations (Figure 5) and putatively dysfunctional. The typical M. plutonius strains harbor an identical enhancin protein (744 amino acids), whereas the enhancin of atypical strain DAT561 is slightly truncated (728 amino acids).

Figure 5: Comparison of the enhancin gene cluster of M. plutonius S1 with P. larvae DSM 25719, P. larvae DSM 25430 and B. thuringiensis serovar kurstaki str. YBT1520. The graphical presentation was done with the Easyfig software (minimum blast hit length of 50 bp) (Sullivan et al., 2011). ORFs depicted as dotted arrows represent pseudogenes. ORFs related to enhancin are orange and transposases are shown in yellow.

Furthermore, we detected a gene encoding putative endo-alpha-N-acetylgalactosaminidase (EC 3.2.1.97) that catalyzes the release of oligosaccharides via hydrolysis of the O-glycosidic bond between alpha-acetylgalactosamine at the reducing end of mucin-type sugar chains (O-glycan) and serine/threonine residues of proteins, which is putatively dysfunctional in the M. plutonius strains 119, 82, 90.0 and ATCC 35311 due to nonsense mutations. As shown in Supplementary Figure 2, the peptidase M60 family protein as well as the endo-alpha-N-acetylgalactosaminidase are transcribed in vivo during EFB pathogenesis, while an expression of these putative virulence factors was not detected in a healthy honey bee larva (data not shown).

3.3.4. Cell surface and adhesion-associated proteins

The ability to adhere to extracellular matrix proteins of animal cells like fibronectin, fibrinogen, collagen, and laminin is the first and critical step to establish an infection for many pathogenic bacteria (Courtney et al., 1994; Holmes et al., 2001; Massey et al., 2001;

Spigaglia et al., 2013). Altogether five gene clusters and three single ORFs were associated with cell surface and adhesion and putatively have an impact on virulence. Each typical strain has nonsense mutations in at least one of the cluster involved in adhesion (Supplementary Data 2, Sheet 4). An overview of the identified cell surface and adhesion-associated proteins including their domain structures is depicted in Figure 6 and the presence and absence of selected proteins is shown in Figure 4 and Supplementary Data 2, Sheet 4.

Interestingly, the genomes of the typical strains encode less potentially functional cell surface and adhesion-associated proteins than the atypical strains DAT561. Two gene clusters (one and five) of the typical strains are putative remnants of clusters detected in the atypical strain DAT561, cluster three contains one ORF with a nonsense mutation and cluster four is missing in all typical strains (Figure 4).

A fibronectin/fibrinogen-binding domain (DUF814)-containing protein was discovered in all strains used in this study (Figure 4 and Figure 6). The corresponding ORF encodes a protein, which shares high similarity (70% identity and 99% coverage) to the fibronectin-binding protein of Enterococcus caccae and E. moraviensis (WP_010772361 and WP_010765067, respectively). Fibronectin and fibrinogen are essential parts of the extracellular matrix of animal cells. Thus, many bacterial pathogens harbor proteins for adhesion involving these proteins, e.g. FbpA, a surface fibronectin-binding protein required for intestinal and liver colonization of Listeria monocytogenes (Dramsi et al., 2004). A putative extracellular matrix-binding protein (MEPL7_19p00060, Figure 4 and Figure 6) is

plasmid-encoded (pMP19) and only present in the typical strains 21.1, 49.3, 60 and H6 (Supplementary Data 2, Sheet 4). It contains eight copies of a DUF1542 domain. In Staphylococcus aureus it was shown that some DUF1542-containing proteins are involved in cell cluster formation, cellular adhesion and antibiotic resistance (Clarke et al., 2002;

Schroeder et al., 2009). This protein shares the highest amino acid sequence identity (47%) with a matrix-binding protein of Lactobacillus rhamnosus (WP_033571521).

Figure 6: Domain structure of putative cell surface and adhesion proteins identified in M. plutonius 49.3 and DAT561 with a putative role as virulence factors. Signal peptides, transmembrane regions and domains were determined using InterProScan 5, and are depicted using the color code shown in the legend. Cluster sizes range from 2 to 5.3 kbp in M. plutonius 49.3 and 1.7 kbp to 9.8 kbp in M.

plutonius DAT561. The presence of orthologous genes and gene cluster identified in the other strains are shown in Figure 4 and Supplementary Data 2, Sheet 4.

3.3.5. Toxin

In addition to M. plutonius, Paenibacillus larvae causes also an important bacterial disease afflicting the honey bee brood (American foulbrood (AFB)). In a recent review, it was shown that P. larvae attacks honey bee larvae epithelial cells via secreted toxins during pathogenesis (Poppinga and Genersch, 2015). These P. larvae-toxins destroy the epithelial integrity and enable bacteria to breach the epithelium of larvae via a paracellular route. Only the genomes of the typical M. plutonius strains 21.1, 49.3, 60 and H6 harbor a putative toxin-encoding ORF (Supplementary Data 2, Sheet 4), while all other typical strains and the atypical strain DAT561 lack a respective gene. The toxin, now designated “melissotoxin A”, is plasmid-encoded (pMP19). It shows 33% amino acid sequence identity to an epsilon toxin ETX/mosquitocidal toxin MTX2 family protein of Brevibacillus laterosporus (WP_018669999), a common secondary invader in EFB disease (Djukic et al., 2011). The putative toxin harbors an N-terminal signal peptide and is similar to proteins of PFAM family PF03318 such as the Clostridium epsilon toxin ETX and the Bacillus mosquitocidal toxin MTX2, and Aerolysin-like toxins. Noteworthy, the melissotoxin A-encoding gene is expressed during infection in vivo (Supplementary Figure 2).

3.3.6. Capsule/cell envelope-forming proteins

Capsules are a layer of surface-associated polysaccharides. They protect the bacteria against desiccation, attack from phages, antimicrobial peptides, and sometimes from phagocytosis (Schembri et al., 2004; Campos et al., 2004). We detected four gene clusters, which are associated with capsule and cell envelope-forming proteins (Supplementary Data 2, Sheet 4).

Gene cluster 1 comprises a putative capsule locus, which was described for E. faecium strains by Palmer et al. (2012). The putative capsule-encoding gene cluster of E. faecium 504 and E. caccae ATCC BAA-1240 share high sequence similarity to this cluster, although all Melissococcus strains contain nonsense mutations in genes involved in capsule formation (Supplementary Data 2, Sheet 4). The number of putatively non-functional ORFs due to nonsense mutations varies between one (strain DAT561) and three to four (all typical strains).

The second gene cluster has a similar composition as the enterococcal polysaccharide antigen (epa)-locus of E. faecalis (Xu et al., 1997; Hancock et al., 2012; Xu et al., 1998), E.

haemoperoxidus and E. caccae (Figure 7). A role of Epa as virulence factor is suggested at least for E. faecalis (Hancock et al., 2012). Furthermore, Epa facilitates resistance to bile salts and antimicrobial peptides (Rigottier-Gois et al., 2014). M. plutonius ATCC 35311 and the atypical strain DAT561 are the only strains, which have frameshift mutations in at least one gene of this cluster (Supplementary Data 2, Sheet 4).

Cluster three and four consist of two ORFs each, which both are putatively only functional in the atypical strain DAT561 (Supplementary Data 2, Sheet 4). ORFs belonging to these clusters encode lipid A-like transporters.

Figure 7: Comparison of a gene cluster of M. plutonius with gene clusters of Enterococci encoding for Epa. ORFs labeled with locus tags represent the corresponding ends of the shown genome segments. ORFs related to the epa-locus are marked in orange and ORFs encoding Epa of E. faecalis are depicted in red. Conserved hypothetical proteins are shown in gray. The gene cluster shows highest sequence similarity to Epa of E. faecalis and E. haemoperoxidus.